Lecture 11

Problem 64. Let (π,V)(\pi,V) be a finite-dimensional representation of a finite group GG. Define a representation (π,V)(\pi^{*},V^{*}) by

[π(g)λ](v):=λ(π(g1)v)gG,λV,vV.[\pi^{*}(g)\lambda](v):=\lambda\big{(}\pi(g^{-1})v\big{)}\qquad\forall g\in G,% \lambda\in V^{*},\,v\in V.
  1. (a)

    Let 𝒜={λi}i=1,,dimV{\mathcal{A}}=\{\lambda_{i}\}_{i=1,\ldots,\operatorname{dim}V^{*}} be a basis of VV^{*} and ={vi}i=1,,dimV{\mathcal{B}}=\{v_{i}\}_{i=1,\ldots,\operatorname{dim}V} the dual basis of VV, i.e.

    λi(vj)={1ifi=j,0ifij..\lambda_{i}(v_{j})=\begin{cases}1\quad&\mathrm{if}\;i=j,\\ 0\quad&\mathrm{if}\;i\neq j.\end{cases}.

    Show that [π(g)]𝒜=[π(g1)]T[\pi^{*}(g)]_{{\mathcal{A}}}=[\pi(g^{-1})]_{{\mathcal{B}}}^{T} and χπ=χπ¯\chi_{\pi^{*}}=\overline{\chi_{\pi}}.

  2. (b)

    Show that (π,V)(\pi^{*},V^{*}) is irreducible if and only if (π,V)(\pi,V) is.

Solution: Let ,\langle\cdot,\cdot\rangle be an inner product on VV under which {vi}\{v_{i}\} is an orthonormal basis. Then

π(g1)vj=π(g1)vj,v1v1+π(g1)vj,v2v2++π(g1)vj,vdvd\pi(g^{-1})v_{j}=\langle\pi(g^{-1})v_{j},v_{1}\rangle v_{1}+\langle\pi(g^{-1})% v_{j},v_{2}\rangle v_{2}+\ldots+\langle\pi(g^{-1})v_{j},v_{d}\rangle v_{d}

(where d=dimVd=\operatorname{dim}V), hence

[π(g1)]=(π(g1)vj,vi)i,j.[\pi(g^{-1})]_{{\mathcal{B}}}=\big{(}\langle\pi(g^{-1})v_{j},v_{i}\rangle\big{% )}_{i,j}.

The matrix of π(g)\pi^{*}(g) with respect to the basis 𝒜{\mathcal{A}} is given by

[π(g)]𝒜=([π(g)λ1]𝒜[π(g)λd]𝒜).[\pi^{*}(g)]_{{\mathcal{A}}}=\left(\begin{matrix}\vdots&&\vdots\\ [\pi^{*}(g)\lambda_{1}]_{{\mathcal{A}}}&\cdots&[\pi^{*}(g)\lambda_{d}]_{{% \mathcal{A}}}\\ \vdots&&\vdots\end{matrix}\right).

Writing [π(g)λj]=a1jλ1+a2jλ2++adjλd[\pi^{*}(g)\lambda_{j}]=a_{1j}\lambda_{1}+a_{2j}\lambda_{2}+\ldots+a_{dj}% \lambda_{d}, we have that

[π(g)λj]𝒜=(a1ja2jadj),[\pi^{*}(g)\lambda_{j}]_{{\mathcal{A}}}=\left(\begin{matrix}a_{1j}\\ a_{2j}\\ \vdots\\ a_{dj}\end{matrix}\right),

thus [π(g)]𝒜=(aij)i,j[\pi^{*}(g)]_{{\mathcal{A}}}=\big{(}a_{ij}\big{)}_{i,j}. We now observe that

[π(g)λj](vi)=a1jλ1(vi)+a2jλ2(vi)++adjλd(vi)=aij.[\pi^{*}(g)\lambda_{j}](v_{i})=a_{1j}\lambda_{1}(v_{i})+a_{2j}\lambda_{2}(v_{i% })+\ldots+a_{dj}\lambda_{d}(v_{i})=a_{ij}.

On the other hand, from the definition of π\pi^{*}, we have

[π(g)λj](vi)=λj(π(g1)vi)\displaystyle[\pi^{*}(g)\lambda_{j}](v_{i})=\lambda_{j}\big{(}\pi(g^{-1})v_{i}% \big{)} =λj(π(g1)vi,v1v1+π(g1)vi,v2v2++π(g1)vi,vdvd)\displaystyle=\lambda_{j}\left(\langle\pi(g^{-1})v_{i},v_{1}\rangle v_{1}+% \langle\pi(g^{-1})v_{i},v_{2}\rangle v_{2}+\ldots+\langle\pi(g^{-1})v_{i},v_{d% }\rangle v_{d}\right)
=π(g1)vi,vj,\displaystyle=\langle\pi(g^{-1})v_{i},v_{j}\rangle,

giving aij=π(g1)vi,vja_{ij}=\langle\pi(g^{-1})v_{i},v_{j}\rangle, hence

[π(g)]𝒜=(aij)ij=(π(g1)vi,vj)ij=(π(g1)vj,vi)ijT=[π(g1)]T.[\pi^{*}(g)]_{{\mathcal{A}}}=\big{(}a_{ij}\big{)}_{ij}=\big{(}\langle\pi(g^{-1% })v_{i},v_{j}\rangle\big{)}_{ij}=\big{(}\langle\pi(g^{-1})v_{j},v_{i}\rangle% \big{)}_{ij}^{T}=[\pi(g^{-1})]_{{\mathcal{B}}}^{T}.

Using this, we compute:

χπ(g)=tr[π(g)]𝒜=tr([π(g1)]T)=tr([π(g1)])=χπ(g1)=χπ(g)¯.\chi_{\pi^{*}}(g)=\operatorname{tr}[\pi^{*}(g)]_{{\mathcal{A}}}=\operatorname{% tr}\big{(}[\pi(g^{-1})]_{{\mathcal{B}}}^{T}\big{)}=\operatorname{tr}\big{(}[% \pi(g^{-1})]_{{\mathcal{B}}}\big{)}=\chi_{\pi}(g^{-1})=\overline{\chi_{\pi}(g)}.

This gives χπG=χπG\|\chi_{\pi}\|_{G}=\|\chi_{\pi^{*}}\|_{G}, hence (π,V)(\pi,V) is irreducible (χπG=1\|\chi_{\pi}\|_{G}=1) iff (V,π)(V^{*},\pi^{*}) is (χπG=1\|\chi_{\pi^{*}}\|_{G}=1).

Problem 65. Let (π,V)(\pi,V) be a representation of a group GG and (π,W)(\pi,W) a subrepresentation of (π,V)(\pi,V). Show that (π~,V/W)(\widetilde{\pi},V/W) is a representation of GG, where

π~(g)(v+W):=π(g)v+Wv+WV/W.\widetilde{\pi}(g)(v+W):=\pi(g)v+W\qquad\forall v+W\in V/W.

Solution: We first check that π~\widetilde{\pi} is well-defined: given vVv\in V, wWw\in W, we have

π(g)(v+w)=π(g)v+π(g)w,\pi(g)(v+w)=\pi(g)v+\pi(g)w,

and since (π,W)(\pi,W) is a subrepresentation of (π,V)(\pi,V), π(g)wW\pi(g)w\in W. This means that for each wWw\in W, π(g)(v+w)\pi(g)(v+w) gets mapped to π(g)v+W\pi(g)v+W under the quotient map VV/WV\rightarrow V/W. Hence π~(g):V/WV/W\widetilde{\pi}(g):V/W\rightarrow V/W is a well-defined linear map for each gGg\in G. We now verify that π~:GGL(V/W)\widetilde{\pi}:G\rightarrow\operatorname{GL}(V/W) is a group homomorphism. Firstly,

π~(e)(v+W)=π~(e)v+W=v+W,\widetilde{\pi}(e)(v+W)=\widetilde{\pi}(e)v+W=v+W,

thus π~(e)=Id\widetilde{\pi}(e)={\,\mathrm{Id}}. We also verify the homomorphism property:

π~(gh)(v+W)=π(gh)v+W=π(g)π(h)v+W=π~(g)(π(h)v+W)=π~(g)(π~(h)(v+W)),\widetilde{\pi}(gh)(v+W)=\pi(gh)v+W=\pi(g)\pi(h)v+W=\widetilde{\pi}(g)(\pi(h)v% +W)=\widetilde{\pi}(g)\big{(}\widetilde{\pi}(h)(v+W)\big{)},

i.e. π~(gh)=π~(g)π~(h)\widetilde{\pi}(gh)=\widetilde{\pi}(g)\widetilde{\pi}(h).

Problem 66. Let (π,W)(\pi,W) and (π,W)(\pi,W^{\prime}) be two subrepresentations of a representation (π,V)(\pi,V) of a group GG such that

(π,V)=(π,W)(π,W).(\pi,V)=(\pi,W)\oplus(\pi,W^{\prime}).
  1. (a)

    Show that

    (π~,V/W)(π,W).(\widetilde{\pi},V/W)\cong(\pi,W^{\prime}).
  2. (b)

    Show

    χ(π,V/W)=χ(π,V)χ(π,W).\chi_{(\pi,V/W)}=\chi_{(\pi,V)}-\chi_{(\pi,W)}.

Solution:

  1. (a)

    Since V=WWV=W\oplus W^{\prime}, given vVv\in V, there exist unique wW,wWw^{\prime}\in W^{\prime},w\in W such that v=w+wv=w^{\prime}+w. Thus,

    v+W=(w+w)+W=w+W;v+W=(w^{\prime}+w)+W=w^{\prime}+W;

    and for any uv+Wu\in v+W, u=w+w1u=w^{\prime}+w_{1}, so the map v+Wwv+W\mapsto w^{\prime} is a well-defined invertible linear map from V/WV/W to WW^{\prime}. We show that this intertwines with the GG-actions on V/WV/W and WW^{\prime}. We have

    π(g)v=π(g)(w+w)=π(g)w+π(g)w.\pi(g)v=\pi(g)(w^{\prime}+w)=\pi(g)w^{\prime}+\pi(g)w.

    Since (π,W)(\pi,W^{\prime}) and (π,W)(\pi,W) are subrepresentations of (π,V)(\pi,V), π(g)wW\pi(g)w^{\prime}\in W^{\prime}, π(g)wW\pi(g)w\in W, hence π(g)w+π(g)w\pi(g)w^{\prime}+\pi(g)w is the decomposition of π(g)v\pi(g)v according to WWW^{\prime}\oplus W. Thus, π~(g)(v+W)=π~(g)(w+W)=π(g)w+Wπ(g)w\widetilde{\pi}(g)(v+W)=\widetilde{\pi}(g)(w^{\prime}+W)=\pi(g)w^{\prime}+W% \mapsto\pi(g)w^{\prime} under the map above. This shows that v+Wwv+W\mapsto w^{\prime} is a GG-isomorphism.

  2. (b)

    By Theorem 5.12 we can write (π,V)=(π|W,W)(π|W,W)(\pi,V)=(\pi|_{W},W)\oplus(\pi|_{W^{\prime}},W^{\prime}). Let 𝐰1,𝐰n{\bf w}_{1},\ldots{\bf w}_{n} be a basis of WW and 𝐰1,𝐰m{\bf w}^{\prime}_{1},\ldots{\bf w}^{\prime}_{m} be a basis of WW^{\prime}. Then 𝐰1,𝐰n,𝐰1,𝐰m{\bf w}_{1},\ldots{\bf w}_{n},{\bf w}^{\prime}_{1},\ldots{\bf w}^{\prime}_{m} is a basis of VV. For any gGg\in G we therefore have that

    π(g)=(π|W(g)00π|W(g))\pi(g)=\begin{pmatrix}\pi|_{W}(g)&0\\ 0&\pi|_{W^{\prime}}(g)\end{pmatrix}

    with respect to this basis and thus tr(π(g))=tr(π|W(g))+tr(π|W(g))\operatorname{tr}(\pi(g))=\operatorname{tr}(\pi|_{W}(g))+\operatorname{tr}(\pi% |_{W^{\prime}}(g)) so χπ(g)=χπ|W(g)+χπ|W(g)\chi_{\pi}(g)=\chi_{\pi|_{W}}(g)+\chi_{\pi|_{W^{\prime}}}(g). By Proposition 11.1 we have that (π^,V/W)(π|W,W)(\hat{\pi},V/W)\cong(\pi|_{W^{\prime}},W^{\prime}) so χπ^=χπ|W(g)\chi_{\hat{\pi}}=\chi_{\pi|_{W^{\prime}}}(g) and then it is just a case of substituting this into the previous expression and rearranging.

Problem 67. Consider the representation of S3S_{3} on (ρ,V)(\rho,V), where V=2V={\mathbb{C}}^{2}, given by

ρ(1 2)=(0110),ρ(1 2 3)=(ω00ω2)\rho(1\;2)=\begin{pmatrix}0&1\\ 1&0\end{pmatrix},\qquad\rho(1\;2\;3)=\begin{pmatrix}\omega&0\\ 0&\omega^{2}\end{pmatrix}

with ω=e2πi/3\omega=e^{2\pi i/3}.

  1. (a)

    Write down the matrices of (ρρ)(1 2)(\rho\otimes\rho)(1\;2) and (ρρ)(1 2 3)(\rho\otimes\rho)(1\;2\;3) with respect to the basis

    e1e1,e1e2,e2e1,e2e2e_{1}\otimes e_{1},e_{1}\otimes e_{2},e_{2}\otimes e_{1},e_{2}\otimes e_{2}

    of VVV\otimes V (where e1e_{1} and e2e_{2} are the standard basis of VV).

  2. (b)

    Write the character of VVV\otimes V as a sum of irreducible characters.

  3. (c)

    For each of the irreducible characters of VVV\otimes V used in the previous part, find a subrepresentation of VVV\otimes V with that character.

  4. (d)

    Find an isomorphism from (sgn,)(ρ,V)(\mathrm{sgn},{\mathbb{C}})\otimes(\rho,V) to (ρ,V)(\rho,V).

  5. (e)

    Find an isomorphism from (ρ,V)(\rho^{*},V^{*}) to (ρ,V)(\rho,V).

  6. (f)

    Let (ρn,Vn)=(ρ,V)(ρ,V)(\rho^{\otimes n},V^{\otimes n})=(\rho,V)\otimes\ldots\otimes(\rho,V) with nn factors. Decompose (ρn,Vn)(\rho^{\otimes n},V^{\otimes n}) into irreducible representations.

Solution:

  1. (a)

    We compute:

    ρρ(1 2)(e1e1)=ρ(1 2)e1ρ(1 2)e1=e2e2\displaystyle\rho\otimes\rho(1\;2)(e_{1}\otimes e_{1})=\rho(1\;2)e_{1}\otimes% \rho(1\;2)e_{1}=e_{2}\otimes e_{2}
    ρρ(1 2)(e1e2)=ρ(1 2)e1ρ(1 2)e2=e2e1\displaystyle\rho\otimes\rho(1\;2)(e_{1}\otimes e_{2})=\rho(1\;2)e_{1}\otimes% \rho(1\;2)e_{2}=e_{2}\otimes e_{1}
    ρρ(1 2)(e2e1)=ρ(1 2)e2ρ(1 2)e1=e1e2\displaystyle\rho\otimes\rho(1\;2)(e_{2}\otimes e_{1})=\rho(1\;2)e_{2}\otimes% \rho(1\;2)e_{1}=e_{1}\otimes e_{2}
    ρρ(1 2)(e2e2)=ρ(1 2)e2ρ(1 2)e2=e1e1,\displaystyle\rho\otimes\rho(1\;2)(e_{2}\otimes e_{2})=\rho(1\;2)e_{2}\otimes% \rho(1\;2)e_{2}=e_{1}\otimes e_{1},

    hence

    ρρ(1 2)=(0001001001001000).\rho\otimes\rho(1\;2)=\left(\begin{matrix}0&0&0&1\\ 0&0&1&0\\ 0&1&0&0\\ 1&0&0&0\end{matrix}\right).

    Similarly,

    ρρ(1 2 3)(e1e1)=ρ(1 2 3)e1ρ(1 2 3)e1=ωe1ωe1=ω2(e1e1)\displaystyle\rho\otimes\rho(1\;2\;3)(e_{1}\otimes e_{1})=\rho(1\;2\;3)e_{1}% \otimes\rho(1\;2\;3)e_{1}=\omega e_{1}\otimes\omega e_{1}=\omega^{2}(e_{1}% \otimes e_{1})
    ρρ(1 2 3)(e1e2)=ρ(1 2 3)e1ρ(1 2 3)e2=ωe1ω2e2=e1e2\displaystyle\rho\otimes\rho(1\;2\;3)(e_{1}\otimes e_{2})=\rho(1\;2\;3)e_{1}% \otimes\rho(1\;2\;3)e_{2}=\omega e_{1}\otimes\omega^{2}e_{2}=e_{1}\otimes e_{2}
    ρρ(1 2 3)(e2e1)=ρ(1 2 3)e2ρ(1 2 3)e1=ω2e2ωe1=e2e1\displaystyle\rho\otimes\rho(1\;2\;3)(e_{2}\otimes e_{1})=\rho(1\;2\;3)e_{2}% \otimes\rho(1\;2\;3)e_{1}=\omega^{2}e_{2}\otimes\omega e_{1}=e_{2}\otimes e_{1}
    ρρ(1 2 3)(e2e2)=ρ(1 2 3)e2ρ(1 2 3)e2=ω2e2ω2e2=ω(e2e2),\displaystyle\rho\otimes\rho(1\;2\;3)(e_{2}\otimes e_{2})=\rho(1\;2\;3)e_{2}% \otimes\rho(1\;2\;3)e_{2}=\omega^{2}e_{2}\otimes\omega^{2}e_{2}=\omega(e_{2}% \otimes e_{2}),

    hence

    ρρ(1 2 3)=(ω200001000010000ω).\rho\otimes\rho(1\;2\;3)=\left(\begin{matrix}\omega^{2}&0&0&0\\ 0&1&0&0\\ 0&0&1&0\\ 0&0&0&\omega\end{matrix}\right).
  2. (b)

    Recall the character table of S3S_{3}:

    ee (1 2)(1\;2) (1 2 3)(1\;2\;3)
    (Id,)({\,\mathrm{Id}},{\mathbb{C}}) 1 1 1
    (sgn,)(\mathrm{sgn},{\mathbb{C}}) 1 -1 1
    (π,W0)(\pi,W_{0}), 2 0 -1

    Using (a), we compute the character of ρρ\rho\otimes\rho:

    ee (1 2)(1\;2) (1 2 3)(1\;2\;3)
    ρρ\rho\otimes\rho 4 0 1

    Either by using the inner product on functions on S3S_{3}, or by direct inspection of the character table, we see that

    χρρ=χId+χsgn+χπ.\chi_{\rho\otimes\rho}=\chi_{{\,\mathrm{Id}}}+\chi_{\mathrm{sgn}}+\chi_{\pi}.
  3. (c)

    The computations in part (a) tell us that span{e1e2,e2e1}\mathrm{span}\{e_{1}\otimes e_{2},e_{2}\otimes e_{1}\} and span{e1e1,e2e2}\mathrm{span}\{e_{1}\otimes e_{1},e_{2}\otimes e_{2}\} are invariant subspaces of VVV\otimes V. Futhermore,

    ρρ(1 2)(e1e2+e2e1)=e1e2+e2e1,ρρ(1 2 3)(e1e2+e2e1)=e1e2+e2e1,\rho\otimes\rho(1\;2)(e_{1}\otimes e_{2}+e_{2}\otimes e_{1})=e_{1}\otimes e_{2% }+e_{2}\otimes e_{1},\quad\rho\otimes\rho(1\;2\;3)(e_{1}\otimes e_{2}+e_{2}% \otimes e_{1})=e_{1}\otimes e_{2}+e_{2}\otimes e_{1},

    hence (ρρ,(e1e2+e2e1))(Id,)\big{(}\rho\otimes\rho,{\mathbb{C}}(e_{1}\otimes e_{2}+e_{2}\otimes e_{1})\big% {)}\cong({\,\mathrm{Id}},{\mathbb{C}}). Similarly,

    ρρ(1 2)(e1e2e2e1)=e2e1e1e2,ρρ(1 2 3)(e1e2e2e1)=e1e2e2e1,\rho\otimes\rho(1\;2)(e_{1}\otimes e_{2}-e_{2}\otimes e_{1})=e_{2}\otimes e_{1% }-e_{1}\otimes e_{2},\quad\rho\otimes\rho(1\;2\;3)(e_{1}\otimes e_{2}-e_{2}% \otimes e_{1})=e_{1}\otimes e_{2}-e_{2}\otimes e_{1},

    hence (ρρ,(e1e2e2e1))(sgn,)\big{(}\rho\otimes\rho,{\mathbb{C}}(e_{1}\otimes e_{2}-e_{2}\otimes e_{1})\big% {)}\cong(\mathrm{sgn},{\mathbb{C}}). From the decomposition of VVV\otimes V into irreducibles, we must then have (ρρ,span{e1e1,e2e2})(π,W0)\big{(}\rho\otimes\rho,\mathrm{span}\{e_{1}\otimes e_{1},e_{2}\otimes e_{2}\}% \big{)}\cong(\pi,W_{0}).

  4. (d)

    The GG-isomorphism is defined via T(λ(xy))=(λxλy)T\big{(}\lambda\otimes(\begin{smallmatrix}x\\ y\end{smallmatrix})\big{)}=(\begin{smallmatrix}\lambda x\\ -\lambda y\end{smallmatrix}) on pure tensors λ(xy)\lambda\otimes(\begin{smallmatrix}x\\ y\end{smallmatrix}). One directly verifies the isomorphism property. (Note that since both representations are irreducible, the space of isomorphisms is one-dimensional. The key to finding the isomorphism is to consider the eigenvectors of ρ(1 2)\rho(1\;2); ρ(1 2)(11)=(11)\rho(1\;2)(\begin{smallmatrix}1\\ 1\end{smallmatrix})=(\begin{smallmatrix}1\\ 1\end{smallmatrix}), and ρ(1 2)(11)=(11)\rho(1\;2)(\begin{smallmatrix}1\\ -1\end{smallmatrix})=-(\begin{smallmatrix}1\\ -1\end{smallmatrix}). Since sgn(1 2)=1\mathrm{sgn}(1\;2)=-1, the isomorphism must essentially swap (11)(\begin{smallmatrix}1\\ -1\end{smallmatrix}) and (11)(\begin{smallmatrix}1\\ 1\end{smallmatrix}).)

  5. (e)

    Let λ1(e1)=1,λ1(e2)=0\lambda_{1}(e_{1})=1,\,\lambda_{1}(e_{2})=0, λ2(e1)=0\lambda_{2}(e_{1})=0, λ2(e2)=1\lambda_{2}(e_{2})=1. Then {λ1,λ2}\{\lambda_{1},\lambda_{2}\} is a basis of VV^{*}. By Problem 54, the matrices of ρ(1 2)\rho^{*}(1\;2) and ρ(1 2 3)\rho^{*}(1\;2\;3) with respect to this basis are

    ρ(1 2)=(0110),ρ(1 2 3)=(ω200ω).\rho^{*}(1\;2)=\left(\begin{matrix}0&1\\ 1&0\end{matrix}\right),\qquad\rho^{*}(1\;2\;3)=\left(\begin{matrix}\omega^{2}&% 0\\ 0&\omega\end{matrix}\right).

    We see that swapping the order of λ1\lambda_{1} and λ2\lambda_{2} gives the same matrices for ρ\rho^{*} and ρ\rho, hence the linear extension of the map λ1e2\lambda_{1}\mapsto e_{2}, λ2e1\lambda_{2}\mapsto e_{1} provides a GG-isomorphism from VV^{*} to VV.

  6. (f)

    Using the fact that χρn=χρn\chi_{\rho^{\otimes n}}=\chi_{\rho}^{n}, we have

    ee (1 2)(1\;2) (1 2 3)(1\;2\;3)
    ρn\rho^{\otimes n} 4n4^{n} 0 1

    We now compute the inner product of χρn\chi_{\rho^{\otimes n}} with the irreducible characters:

    χρn,χIdS3=16(14n+30+21)=4n+26,\langle\chi_{\rho^{\otimes n}},\chi_{{\,\mathrm{Id}}}\rangle_{S_{3}}={% \textstyle\frac{1}{6}}\big{(}1\cdot 4^{n}+3\cdot 0+2\cdot 1\big{)}={\textstyle% \frac{4^{n}+2}{6}},
    χρn,χsgnS3=16(14n1+30(1)+211)=4n+26,\langle\chi_{\rho^{\otimes n}},\chi_{\mathrm{sgn}}\rangle_{S_{3}}={\textstyle% \frac{1}{6}}\big{(}1\cdot 4^{n}\cdot 1+3\cdot 0\cdot(-1)+2\cdot 1\cdot 1\big{)% }={\textstyle\frac{4^{n}+2}{6}},
    χρn,χπS3=16(14n2+300+21(1))=4n13.\langle\chi_{\rho^{\otimes n}},\chi_{\pi}\rangle_{S_{3}}={\textstyle\frac{1}{6% }}\big{(}1\cdot 4^{n}\cdot 2+3\cdot 0\cdot 0+2\cdot 1\cdot(-1)\big{)}={% \textstyle\frac{4^{n}-1}{3}}.

    This gives

    (ρn,Vn)(Id,)4n+26(sgn,)4n+26(π,W0)4n13.\big{(}\rho^{\otimes n},V^{\otimes n}\big{)}\cong({\,\mathrm{Id}},{\mathbb{C}}% )^{\oplus{\textstyle\frac{4^{n}+2}{6}}}\oplus(\mathrm{sgn},{\mathbb{C}})^{% \oplus{\textstyle\frac{4^{n}+2}{6}}}\oplus(\pi,W_{0})^{\oplus{\textstyle\frac{% 4^{n}-1}{3}}}.

Problem 68. Let (π,V)(\pi,V), (ρ,U)(\rho,U), and (σ,W)(\sigma,W) be three finite-dimensional representations of a finite group GG. Show that

  1. (a)

    ((π,V)(ρ,U))(σ,W)(π,V)((ρ,U)(σ,W))\big{(}(\pi,V)\otimes(\rho,U)\big{)}\otimes(\sigma,W)\cong(\pi,V)\otimes\big{(% }(\rho,U)\otimes(\sigma,W)\big{)}

  2. (b)

    ((π,V)(ρ,U))(σ,W)((ρ,U)(σ,W))((π,V)(σ,W))\big{(}(\pi,V)\oplus(\rho,U)\big{)}\otimes(\sigma,W)\cong\big{(}(\rho,U)% \otimes(\sigma,W)\big{)}\oplus\big{(}(\pi,V)\otimes(\sigma,W)\big{)}

For more of a challenge, prove these without the assumptions that the representations are finite-dimensional and the group is finite!
Remark:
Note that (a) lets us write (π,V)(ρ,U)(σ,W)(\pi,V)\otimes(\rho,U)\otimes(\sigma,W) unambiguously (and similarly for larger products of representations).

Solution: For finite GG and finite-dimensional representations π,ρ,σ\pi,\rho,\sigma, we may simply appeal to Proposition 35 and Theorem 29: the character of the left-hand side of (a) equals

χ(πρ)σ=χπρχσ=χπχρχσ=χπ(χρχσ)=χπχρσ=χπ(ρσ),\chi_{(\pi\otimes\rho)\otimes\sigma}=\chi_{\pi\otimes\rho}\chi_{\sigma}=\chi_{% \pi}\chi_{\rho}\chi_{\sigma}=\chi_{\pi}\big{(}\chi_{\rho}\chi_{\sigma}\big{)}=% \chi_{\pi}\chi_{\rho\otimes\sigma}=\chi_{\pi\otimes(\rho\otimes\sigma)},

proving the two representations are isomorphic. For (b) we also use Proposition 17:

χ(πρ)σ=χπρχσ=(χπ+χρ)χσ=χπχσ+χρχσ=χπσ+χρσ=χ(πσ)(ρσ).\chi_{(\pi\oplus\rho)\otimes\sigma}=\chi_{\pi\oplus\rho}\chi_{\sigma}=(\chi_{% \pi}+\chi_{\rho})\chi_{\sigma}=\chi_{\pi}\chi_{\sigma}+\chi_{\rho}\chi_{\sigma% }=\chi_{\pi\otimes\sigma}+\chi_{\rho\otimes\sigma}=\chi_{(\pi\otimes\sigma)% \oplus(\rho\otimes\sigma)}.

More generally, consider the linear maps defined on pure tensors by

(vu)w(vu)w,(v+u)wvw+uw.(v\otimes u)\otimes w\mapsto(v\otimes u)\otimes w,\quad(v+u)\otimes w\mapsto v% \otimes w+u\otimes w.

As usual, one can directly verify that these are GG-isomorphisms between the representations.

Problem 69. Let e,fe,f form a basis of 2{\mathbb{C}}^{2}. Show that ee+ff22e\otimes e+f\otimes f\in{\mathbb{C}}^{2}\otimes{\mathbb{C}}^{2} cannot be written in the form vw22v\otimes w\in{\mathbb{C}}^{2}\otimes{\mathbb{C}}^{2}.

Solution: Assuming ee+ff=vwe\otimes e+f\otimes f=v\otimes w, we write v=ae+bfv=ae+bf and w=ce+dfw=ce+df. Then

ee+ff=vw=(ae+bf)(ce+df)=ac(ee)+ad(ef)+bc(fe)+bd(ff).e\otimes e+f\otimes f=v\otimes w=(ae+bf)\otimes(ce+df)=ac(e\otimes e)+ad(e% \otimes f)+bc(f\otimes e)+bd(f\otimes f).

By Proposition 11.3, ee,ef,fe,ffe\otimes e,e\otimes f,f\otimes e,f\otimes f is a basis of 22{\mathbb{C}}^{2}\otimes{\mathbb{C}}^{2}, hence ac=bd=1ac=bd=1, ad=bc=0ad=bc=0. But this system of equations has no solution!

Problem 70. For two finite-dimensional representations (π,V)(\pi,V) and (ρ,W)(\rho,W) of a finite group GG, show that

(cπρ,Hom(V,W))(π,V)(ρ,W).\big{(}c_{\pi}^{\rho},\operatorname{Hom}(V,W)\big{)}\cong(\pi^{*},V^{*})% \otimes(\rho,W).

Solution: Define

ϕ:VW\displaystyle\phi:V^{*}\otimes W Hom(V,W)\displaystyle\longrightarrow\operatorname{Hom}(V,W)
λ𝐰\displaystyle\lambda\otimes{\bf w} (𝐯λ(𝐯)𝐰).\displaystyle\longmapsto\left({\bf v}\mapsto\lambda({\bf v}){\bf w}\right).

We extend this linearly so we need to show this is a bijection and commutes with the GG-actions. Suppose λ𝐰kerϕ\lambda\otimes{\bf w}\in\ker\phi, then ϕ(λ𝐰)(𝐯)=0\phi(\lambda\otimes{\bf w})({\bf v})=0 and we note that this is only the case when λ=0\lambda=0 or 𝐰=0{\bf w}=0. Hence kerϕ\ker\phi is trivial. Furthermore, by Proposition 11.3, dim(VW)=dim(V)dim(W)=dim(V)dim(W)=dim(Hom(V,W))\operatorname{dim}(V^{*}\otimes W)=\operatorname{dim}(V^{*})\operatorname{dim}% (W)=\operatorname{dim}(V)\operatorname{dim}(W)=\operatorname{dim}(% \operatorname{Hom}(V,W)) so the vector spaces have the same dimension and ϕ\phi is a bijection.

Let gGg\in G be arbitrary, then for any λ𝐰VW\lambda\otimes{\bf w}\in V^{*}\otimes W we have

cπρ(g)(ϕ(λ𝐰))\displaystyle c_{\pi}^{\rho}(g)\left(\phi(\lambda\otimes{\bf w})\right) =cπρ(g)(𝐯λ(𝐯)𝐰)\displaystyle=c_{\pi}^{\rho}(g)\left({\bf v}\mapsto\lambda({\bf v}){\bf w}\right)
=𝐯ρ(g)λ(π(g1)𝐯)𝐰\displaystyle={\bf v}\mapsto\rho(g)\lambda(\pi(g^{-1}){\bf v}){\bf w}
=λ(π(g1)𝐯)ρ(g)𝐰\displaystyle=\lambda(\pi(g^{-1}){\bf v})\rho(g){\bf w}
=ϕ((vλ(π(g1)𝐯))ρ(g)𝐰)=ϕ((π(g)ρ(g))(λ𝐰)).\displaystyle=\phi\left(\left(v\mapsto\lambda(\pi(g^{-1}){\bf v})\right)% \otimes\rho(g){\bf w}\right)=\phi\left(\left(\pi^{*}(g)\otimes\rho(g)\right)(% \lambda\otimes{\bf w})\right).

Thus ϕ\phi commutes with the GG-actions and is a GG-isomorphism.

Problem 71. Given a finite group GG, let (π,(G))\big{(}\pi,{\mathbb{C}}(G)\big{)} denote the permutation representation of GG on (G){\mathbb{C}}(G) associated to the conjugation action of GG on itself, i.e.

π(g)(hGzhh)=hGzhghg1=hGzg1hghgG,hGzhh(G).\pi(g)\left(\sum_{h\in G}z_{h}h\right)=\sum_{h\in G}z_{h}ghg^{-1}=\sum_{h\in G% }z_{g^{-1}hg}h\qquad\forall g\in G,\;\sum_{h\in G}z_{h}h\in{\mathbb{C}}(G).
  1. (a)

    Show that χπ(g)=|G||𝒞g|\chi_{\pi}(g)=\frac{|G|}{|{\mathcal{C}}_{g}|}, where 𝒞g{\mathcal{C}}_{g} is the conjugacy class gg belongs to. Hint: use Problem 46.

  2. (b)

    Then show that

    (π,(G))ρIrr(G)(ρ,Wρ)(ρ,Wρ).\big{(}\pi,{\mathbb{C}}(G)\big{)}\cong\bigoplus_{\rho\in\mathrm{Irr}(G)}(\rho,% W_{\rho})\otimes(\rho^{*},W^{*}_{\rho}).

Solution:

  1. (a)

    First note that under this conjugation action the orbit of gg is its conjugacy class 𝒞g{\mathcal{C}}_{g} and the stabiliser is the set {hG|hgh1=g}\{h\in G\,|\,hgh^{-1}=g\}, for all gGg\in G. By Problem 46 we have χπ(g)=#{hG|ghg1=h}\chi_{\pi}(g)=\#\{h\in G\,|\,ghg^{-1}=h\}. We note that ghg1=hghg^{-1}=h is equivalent to hgh1=ghgh^{-1}=g so χπ(g)\chi_{\pi}(g) is the order of the stabiliser of gg under the conjugation action. Thus by the orbit stabiliser theorem χπ(g)=|G|𝒞g\chi_{\pi}(g)=\frac{|G|}{{\mathcal{C}}_{g}}.

  2. (b)

    First note that we can write the character of ρIrr(G)(ρ,Wρ)(ρ,Wρ)\bigoplus_{\rho\in\mathrm{Irr}(G)}(\rho,W_{\rho})\otimes(\rho^{*},W^{*}_{\rho}), which we denote as χ\chi as

    χ(g)=ρIrr(G)χρ(g)χρ(g),\chi(g)=\sum_{\rho\in\mathrm{Irr}(G)}\chi_{\rho}(g)\chi_{\rho^{*}}(g),

    for all gGg\in G, by using Propositions 7.9(iii) and 11.6. Then we note that by Lemma 9.2 we have

    χ(g)=ρIrr(G)χρ(g)χρ(g)¯,\chi(g)=\sum_{\rho\in\mathrm{Irr}(G)}\chi_{\rho}(g)\overline{\chi_{\rho}(g)},

    and as the sum is over all irreducibles this is the dot product of a column of the character table with itself, hence by Corollary 8.7 χ(g)=|G|𝒞g\chi(g)=\frac{|G|}{{\mathcal{C}}_{g}}, for all gGg\in G. Then as both representations have the same character values for all gGg\in G we have that they are isomorphic, by Lemma 7.5.